[CPMD-list] about oxygen Ultra-soft PP generated by USPP

[Axel Kohlmeyer]

On Mon, 2 Apr 2007, qfzhang wrote:

QZ> Hi,
QZ>   Thanks for the reply! the input file is attached here.I use MESH 
QZ> and DUAL, and

why MESH???

QZ> calculate a charged system. Is there some other keyword that 
QZ> should be specified?

QZ>  I have tried to add l=2 to generate O US-PP.But it shows that 
QZ>  ' no tuning point fot state nnlz = 320 on iteration 1' what's
QZ> wrong with it? The files were also attached here.

as i was alluding to in the previous mail. this state is not bound.
please check the literature on generating pseudopotentials...

QZ>    Another question is can I use UPF US-PP? I see there are UPF
QZ> US-PP on the web page of PWscf.How can I change them between the
QZ> UPF's and FORMATTED's?

there is no conversion between UPF and native uspp.
anyways, i would actually primarily suspect the Zr and Ba
pseudopotentials. the parameter set shipped with the vdb
code was seemingly tuned for an older version of the code
which had a different fitting algorithm. with the current
code you get (unphysical) negative densities for both of them,
so i would suspect _them_ before i suspect the d-channel of 
the oxygen having an impact.

cheers,
   axel.
   
   Best wishes
                                                     Qianfan Zhang  
Axel Kohlmeyer д:

> On Mon, 2 Apr 2007, qfzhang wrote:
> 
> QZ> Hi, dear all!
> 
> dear qianfan zhang.
> 
> QZ>   I want to use USPP to generate the US-PP for oxygen in the
> QZ> calculation of oxi de. But the result is not good in the oxide(even
> QZ> I use very large supercell). I want to add l=2 channel in the O
> 
> first of all, if you don\'t tell us, what you are trying to do and
> what your input looks like, i would strongly suspect that you are
> doing something wrong, since the oxygen ultrasofts generated by 
> bernd meyer _do_ work well with a large number of systems, even
> oxides. so please provide some input that _proves_ that the 
> pseudopotential is at fault. thanks.
> 
> QZ> US-PP, because I find that the TM-PP I used bef ore includes the d
> QZ> channel.  But I find it is impossible for me to generate such a
> 
> you should be _very_ careful with that. pseudopotentials created
> with the \'haman\' style PP code that are in the cpmd pp library
> mostly contain a unbound, bogus d channel, which will either have
> little contribution (if used with LMAX=D LOC=P) or may give worse
> results (if used with LMAX=D, which implies LOC=D).
> 
> QZ> US-PP.Is there anyone who know how to do that?
> 
> adding another projector to existing USPP inputs is actually
> rather straightforward. just follow the instructions in the 
> INPUT_XX files that ship with the code. the problem is to have 
> it converged and bound. so you may have to slightly oxidize
> the atom and/or shift some fractional occupation to the d-state.
> 
> QZ>   Or who know some tricks to improve the quality of US-PP?
> 
> before you put too much effort into optimizating the potential,
> you may be better off to stick with a reasonable norm-conserving
> pseudopotential. you avoid some of the overhead and \'only\' need
> some more cpus resources to distribute the work. but then again
> with ultra-softs, CPMD usually does not scale that far, so in case
> you have a large enough resource to run on, you may find NC-PPs 
> about as efficient as US-PPs.
> 
> cheers,
>    axel.
> 
> 
>  
>    Best wishes
>                                                       Qianfan Zhang
>                                                       2007-04-02
> 
> 
> 
> -- 
> =======================================================================
> Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
>    Center for Molecular Modeling   --   University of Pennsylvania
> Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
> tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
> =======================================================================
> If you make something idiot-proof, the universe creates a better idiot.
> 




-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.